Author Topic: Hobbyist RF amp design approach  (Read 5717 times)

0 Members and 1 Guest are viewing this topic.

Offline Teknow

  • Contributor
  • Posts: 33
  • Country: cy
Hobbyist RF amp design approach
« on: May 16, 2017, 09:45:46 am »
I have been searching in a lot of RF design books, application notes, data sheets, and internet sites for a comprehensive and empirical approach to bias RF transistors to be able to design a multistage RF amplifier but still in vain. I miss an approach in reference to certain transistors and their specifications.
These transistors are mainly from Philips (NXP) for example:
BFG520, BFG540, BFR92, BFR93 etc. and AVAGO AT42086 among others.

After some experimenting and learning I could design circuits depending on certain criteria depicted from the data sheet.
Though essential for stability, I intentionally overlook the mathematics concerning stability as it is beyond my comprehension at the moment.  I started learning about impedance matching, smith charts and Sparameters, but sometimes I get stuck in the abstract of confusion.

The approach I adopted (adapted) in designing my circuits is as follows:
1. I refer to the GUM and relevant curves in the data sheet of a transistor and adopt that Ic, Vce, and gain (marginally!).
For example the GUM for the BFR93A is 13dB when Ic=30mA, Vce=8V and f=1GHz.
2. I calculate Ib by dividing Ic on the typical value of Beta from the data sheet.
3. Then I multiply that by 10 for Ib and Ibb.
4. R1= (Vcc-0.65-Ve)/ Ib+Ibb
5. R2= Ve+0.65/ Ibb
6. Re= Ve/ Ic (for a small signal transistor I let Ve= Vce/ 10.
7. Rc= Vce/ Ic.

Imagine I built two identical stages using the BFR93A, and a medium power stage using the BFG135, or the BFG196, or BFG198 etc.

For the BFG198, I follow the above mentioned steps but sometimes I make Ve=0 to get more power and thermal coupling to PCB through the emitter leads.
I use an LM317, or a pnp transistor (active bias) to limit current for this transistor.
After building the three stages and for interstage impedance matching (trial and error) I connect variable capacitors to tweak and monitor for maximum peak on the spectrum analyser.
Once I achieve the best performance , I replace the varicaps with fixed ones.



In order to cultivate this approach for hobbyists and while on the way to simplify Sparameters etc. I need some suggestions and feedback to improve and answers to some questions.
1. In RF design books, the transistor design steps are rarely based on specific RF transistors and their specifications.
2. The beta is "assumed" 50, why not that maximum or the typical!?
3. There are no explicit examples tackling wide-band amps or narrowing amps to varieties of frequencies!
4. There are no subjects tackling trade off of power and thermal stability and how to cool the RF SOT223 transistor!
5. In the Philips medium and high power transistor data-sheets, there are suggested circuits for the intermodulation distortion, can one adopt them as suggested amplifier designs for that transistor (considering the BFG135, ..etc.)
6. These circuits are for specific frequency, how can one modify them?
7. The Vcc and Vbb are never given for these circuits, how can I tell the values?


« Last Edit: May 16, 2017, 10:24:57 am by Teknow »
 
The following users thanked this post: jonovid

Offline dmills

  • Super Contributor
  • ***
  • Posts: 1732
Re: Hobbyist RF amp design approach
« Reply #1 on: May 16, 2017, 08:46:14 pm »
Setting the DC operating point is the same as it is for any other transistor amp, use a beta on the low side in the usual way, higher beta should not seriously impact the Q point if you are using a sane biasing scheme.

Those 47 ohm resistors in series with the tuned collector loads are not doing you any favors, I would get rid or replace with a ferrite bead or such, maybe with 100n to ground from the top of the tank circuits in stages 1 and 2 (All that resistor does is to steal available collector voltage swing).

You might find that a few ohms or so of base resistor is helpful from a stability standpoint.

Base impedance for a common emitter stage is usually very much lower then collector load impedance, a capacitive divider works, an L match is better, or tap a turn or so from the end of the tuned circuit and feed a cap divider from that.

Bipolar biasing for a stage with the emitters strapped hard down (Such as you might do for cooling) is a pain in the arse, think about well thermally coupled current mirrors, but the bias needs to be stiff for low distortion, which is a pain (There is a reason most use fets for this at any power).

EMRFD has some useful notes for the experimenter.

73 Dan.
 

Offline Teknow

  • Contributor
  • Posts: 33
  • Country: cy
Re: Hobbyist RF amp design approach
« Reply #2 on: May 17, 2017, 09:32:32 am »
Thanks a lot Dan! I appreciate your very informative reply.
I checked the reply last night after spending hours in the quest of success while working on similar design to this one. I am going to make the changes you suggested and hope things will be better.
Unfortunately, the circuits I have tried so far barely amplify the input signal (1mW @ 433MHz) or make a difference!
 

Z80

  • Guest
Re: Hobbyist RF amp design approach
« Reply #3 on: May 17, 2017, 10:17:54 am »
Have you read 'RF circuit design' by Chris Bowick?  It is a good book for explaining the basics and addresses most of the questions you raised in a quite approachable way.  Building amplifiers in the high MHz / GHz range is an art, parasitics can kill an otherwise technically sound circuit.
 

Offline dmills

  • Super Contributor
  • ***
  • Posts: 1732
Re: Hobbyist RF amp design approach
« Reply #4 on: May 17, 2017, 04:59:01 pm »
Yep, UHF can be all kinds of fun, but there should be plenty of acceptable 70cms band designs already kicking around.
Up there you will want to do real impedance matching most likely the base impedance will be anything but resistive, or go Pi network with two trimmers.

I think you isolating chokes on that bit of copper clad are almost certainly operating above resonance and are probably doing you no favours.

With 1mW input to a three stage amp, I hope you have a license as that should manage several hundred mW at the output, and you will want decent output filtering as well.
Depending on your modulation, I might make the third stage zero bias class C (Fine for FM, probably ok, more or less, for off on keying, not suitable for QAM). 
 
Regards, Dan.
 

Offline MikeLogix

  • Contributor
  • Posts: 46
  • Country: us
Re: Hobbyist RF amp design approach
« Reply #5 on: May 20, 2017, 09:37:05 am »
Have you read 'RF circuit design' by Chris Bowick?  It is a good book for explaining the basics and addresses most of the questions you raised in a quite approachable way.  Building amplifiers in the high MHz / GHz range is an art, parasitics can kill an otherwise technically sound circuit.

No, it is not a good book, it is Great book for RF design. :)
 

Offline kazam

  • Regular Contributor
  • *
  • Posts: 70
Re: Hobbyist RF amp design approach
« Reply #6 on: May 20, 2017, 01:39:07 pm »
As frequency goes up you will need to understand S-parameters and the Smith chart. At the very least!

I use Keysight ADS for RF power amplifier design and its a good tool, unfortunately not feasible for amateurs.

If you want to change the frequency of your amplifier you need to present a different impedance to the transistor. Or perhaps the same impedance at a different frequency which means your matching network needs to change.

This is a complicated subject and a few rules of thumb simply will not be enough to do anything outside of replicating an already built and tested circuit.
 

Offline cdev

  • Super Contributor
  • ***
  • Posts: 5082
  • Country: 00
Re: Hobbyist RF amp design approach
« Reply #7 on: May 20, 2017, 02:36:08 pm »
RFSIM99 does a basic RF amp and will import s parameters and it's free.


And easy to use.
« Last Edit: May 20, 2017, 07:59:24 pm by cdev »
"What the large print giveth, the small print taketh away."
 

Offline evb149

  • Super Contributor
  • ***
  • Posts: 1666
  • Country: us
Re: Hobbyist RF amp design approach
« Reply #8 on: May 20, 2017, 07:40:53 pm »
Pardon the intrusion, but on that note I have a couple of questions:
* Does anyone know the author's / software's authoritative web site and can share the link?

IIRC I saw the RFSIN99 software cited as being useful somewhere so I went looking for "the official site" and I don't recall that I ever found one.  I'm not even sure I found who the author is or if I did whether the author has any particular web presence.

If I'm recalling correctly I think it might be "abandonware" in the sense that it is no longer maintained which, if true, is sad, but understandable for free software.  But usually even for old free software there is often some ghost of an original web site that has the last distribution and some program notes / documentation such as was available and whatever.

I think I saw a few places claiming to offer downloads of the program but it always makes me uneasy when there's just some "here download this .EXE" link and zero attribution as to its version number / origin / hash / documentation / whatever.  If a mirror is the only source then at least it'd be nice if it was one that was more trustable and which at least said "Here's RFSIM99 v1.14 as retrieved 2011-1-1 from www.authorsite.com and here's the documentation file as of that date".

Beyond that I'd be sort of interested to ask the author if it is abandonware if s/he would not mind to release the source code to the public domain / GPL or such condition so it can at least be inspected, updated a little and maintained.  Maybe it is not so useful compared to modern alternatives to warrant the attention, I don't know. 

Seems like there's a bunch of old stuff like this and maybe Appcad and whatever else that should have a modern equivalent that's open source and cross platform which would probably be easy if original information existed.

Switching to the original question as to biasing and stuff: If I recall correctly it seems like most of the S parameters of these devices are specified at a given VCE and a given IC.  So if you're designing the transfer function and matching based on the S parameters (and I don't think much else could be definitive enough to use for the RF response) then it seems like "there's your answer" -- pick the S-parameters that are closest to the ones you want for your stage understanding the likely "dperformance / dIC" or whatever based on the curves / parameters then design your circuit for that stage to give about that CE voltage and particularly about that IC value when operating.  How exactly you do that seems like it degenerates into literally a few basic (4 or 5 sane options at most?) possible circuits that can DC BIAS a RF BJT with some degree of stability / predictability and not have too much unwanted impact on the RF at the B / E.  Parasitics are of course quite a concern and basically everything touching the B / E / C whether a line segment or PCB mounting pad or the part's package parasitic model itself and the PCB e_r will have to be represented in your SPICE / SPARAM simulation so that you can see the affects on your overall 2 port parameters and stability vs. frequency.

And back to giving a shout out to the likes of APPCAD, IIRC it had some useful utilities in there as well as some of the old HP RF application notes.  Seems like HP / CEL / Motorola had some app notes specifically about biasing and decoupling RF BJTs.  I know I've seen biasing of JFETs/HEMTs but I think also BJTs.

And there are also LNA design application notes for 802.11 applications for some of the NXP BJTs that I think you mentioned right on the product pages that show recommended layout / biasing etc. for their small signal and maybe mid-stage driver parts for that application.  So I can't imagine that those wouldn't be a good starting point for decoupling / biasing.

Motorola's old RF data books / application notes are online and they had lots of examples of medium and higher power circuits as well as small signal stuff for their MRFxxx and other parts and some of those discussed biasing.

From notes:
"Bias Circuits for RF devices" by VA3IUL, on the web.
HP AN 944-1 "Microwave Transistor Bias Considerations"


RFSIM99 does a basic RF amp and will import s parameters and it's free.
And easy to use.
 

Offline cdev

  • Super Contributor
  • ***
  • Posts: 5082
  • Country: 00
Re: Hobbyist RF amp design approach
« Reply #9 on: May 20, 2017, 07:50:52 pm »
You're right, it is abandonware. I got my copy ages ago and simply keep copying it from one machine to another.

I think the site I got it from was this  engineering class home page- affiliated with the University of San Diego but it came with a disclaimer that they were not affiliated with it. The firm that wrote it seems to be long defunct.

Its very simple to use, it does just a few things but it does them well. You can also design simple filters with it.
« Last Edit: May 20, 2017, 08:08:43 pm by cdev »
"What the large print giveth, the small print taketh away."
 
The following users thanked this post: evb149


Offline evb149

  • Super Contributor
  • ***
  • Posts: 1666
  • Country: us
« Last Edit: May 20, 2017, 11:07:51 pm by evb149 »
 
The following users thanked this post: cdev

Offline cdev

  • Super Contributor
  • ***
  • Posts: 5082
  • Country: 00
Re: Hobbyist RF amp design approach
« Reply #12 on: May 20, 2017, 11:50:30 pm »
I have a question, why aren't RF amplifiers cheaper by now? Very small MMICs exist and make amplification for receiver hardware easy to implement. Why aren't there more, similar modular options for transmission?


Or am I missing something?

"What the large print giveth, the small print taketh away."
 

Offline evb149

  • Super Contributor
  • ***
  • Posts: 1666
  • Country: us
Re: Hobbyist RF amp design approach
« Reply #13 on: May 21, 2017, 12:39:14 am »
Hmm well I'm not an expert but it "sort of" seems like there are.
There have been "pre matched" RF power devices for decades that are designed to provide a fairly nice matching impedance and stable operation over some particular band of operation which generally coincides with some commercial service requirement (GSM, whatever). 

So by analogy with a MMIC LNA which is generally "pre matched" to be broadband or at least operable within a given range and matched to 50R a power transistor that is also internally matched gives you "about the same thing".

And medium power amplifiers / drivers are (if I recall correctly) available as modules from places like Mini Circuits. 

You're more likely to want to adjust the match / mismatch with PAs for your custom transmitter design.  And the bias tee / choking / biasing and all that will be more custom to a particular transmitter design and use bulkier components (because of higher power losses / heat concerns etc) and so harder to integrate into a small package so those are usually external.

Then of course you have port inductance and heatsinking so any integrated solution is going to be a large / bulky package and may compromise the performance if one is not careful.

And then there are the variations of class of operation whereas most receiver MMICs will be strictly linear not quite so many transmitters will be biased / designed for that.  And then of course you'd need output harmonic filtering which would be pretty application specific other than for some very generic well defined services like GSM / whatever.

And just look at what's out there in say single LDMOS devices.  They're not that hard to use necessarily without built in "MMIC" type IC design.

And you can buy premade amplifier modules, they're just usually expensive and not some $10 "IC" in low volumes.

So I suppose (AFAIK) the answer is that discrete transistors possibly with some factory engineering to suit their matching / design / packaging to a particular service are considered "good enough" for most applications, particularly cost sensitive ones where adding dollars of extra cost for better included biasing / matching / power supply stuff isn't cost effective vs. a discrete PCBA and "cheap as can be made" ad-hoc stuff.

But yeah it does seem like there's SOME room for "mainstream" power modules that would be pretty "plug and play" for a particular band / service.  In particular when you start to look at things like Class-D/E/F amplifiers or semi-QRP transmitters (you could do that with a standard MMIC pretty much) or ISM stuff or whatever there certainly could be some "ready to go" pre designed / tuned / filtered output amp "packaged devices / modules".

The stuff for amateur HF range stuff is so easy now it is often done with basically SMPS type devices but still quite "tweaky" wrt. tuning filtering and such so making it "cookie cutter" would be nice.  But that's a small market.


I have a question, why aren't RF amplifiers cheaper by now? Very small MMICs exist and make amplification for receiver hardware easy to implement. Why aren't there more, similar modular options for transmission?

Or am I missing something?
 

Offline Teknow

  • Contributor
  • Posts: 33
  • Country: cy
Re: Hobbyist RF amp design approach
« Reply #14 on: May 21, 2017, 08:22:50 am »
Lots of thanks for all contribution to enrich this thread with ideas and knowledge!
Sorry I might not reply to every contribution as I use an archaic IPhone version and super sluggish internet speed and aging eyesight! But I even appreciate all critics!
Now I post those words before I have read the new contributions!
First I assure you there's always a dummy load.
70cm schematics on the Internet is mostly high power modules. I prefer custom "art circuits and designs", and adapting knowledge to my needs and my transistors. Because my Math, knowledge of trigonometry and AC physics are mediocre, it's a must I learn by doing to get along.
Two books are of interest and already on the order list: Sayre's
Complete Wireless Design and Bowick's RF Circuit Design.
I tried all the possible suggestions and impedance match but I had no progress!
I have the feeling that the circuits don't amplify, on the contrary they attenuate! I measure the signal source separately connected to a dummy load, its -60 dB or so. After amplification it's more than -70 dBm!
I think the VCO (signal source) I use produces power in the pico Watt!

Last night I improvised a step up using an ATF21186 followed by MSA0886 (application note 1064) I waste milliamps and I get heat but no amplification! Now I am in the early steps of how to boost a tiny winny signal!
I think beating around the bush of impedance matching is too early and is yet after I learn to boost faint RF signals!
« Last Edit: May 21, 2017, 08:52:02 am by Teknow »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf